derek/TheRock
1
0 Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
Files
1a182e2dac58879f215c0fad0f5f78c76dcafe56
TheRock/compiler/amd-llvm/llvm/test/CodeGen/AMDGPU/GlobalISel/divergence-temporal-divergent-i1.mir
T
derek.holloway 7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
Details
CodeQL / Analyze (python) (push) Canceled after 0s
Details
pre-commit / pre-commit (push) Canceled after 0s
Details
Multi-Arch CI / setup (push) Canceled after 0s
Details
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Details
Setup (Multi-Arch) / setup (push) Canceled after 0s
Details
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Details
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
Details
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Details
Multi-Arch CI / CI Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Details
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Details
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Details
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Details
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
Details
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

769 lines
37 KiB
YAML
Raw Blame History

# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 4
# RUN: llc -mtriple=amdgcn-mesa-amdpal -mcpu=gfx1010 -run-pass=amdgpu-global-isel-divergence-lowering -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX10 %s
---
name: temporal_divergent_i1_phi
legalized: true
tracksRegLiveness: true
body: |
; GFX10-LABEL: name: temporal_divergent_i1_phi
; GFX10: bb.0:
; GFX10-NEXT: successors: %bb.1(0x80000000)
; GFX10-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[C1:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[DEF:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.1:
; GFX10-NEXT: successors: %bb.2(0x04000000), %bb.1(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI:%[0-9]+]]:sreg_32(s1) = PHI [[DEF]](s1), %bb.0, %19(s1), %bb.1
; GFX10-NEXT: [[PHI1:%[0-9]+]]:_(s32) = G_PHI %7(s32), %bb.1, [[C]](s32), %bb.0
; GFX10-NEXT: [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[C]](s32), %bb.0, %9(s32), %bb.1
; GFX10-NEXT: [[PHI3:%[0-9]+]]:_(s1) = G_PHI [[C1]](s1), %bb.0, %11(s1), %bb.1
; GFX10-NEXT: [[COPY3:%[0-9]+]]:sreg_32(s1) = COPY [[PHI3]](s1)
; GFX10-NEXT: [[COPY4:%[0-9]+]]:sreg_32(s1) = COPY [[PHI]](s1)
; GFX10-NEXT: [[C2:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[PHI3]], [[C2]]
; GFX10-NEXT: [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI2]](s32)
; GFX10-NEXT: [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[COPY]]
; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI2]], [[C3]]
; GFX10-NEXT: [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[FCMP]](s1), [[PHI1]](s32)
; GFX10-NEXT: [[S_ANDN2_B32_:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY4]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY3]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_]](s1), [[S_AND_B32_]](s1), implicit-def $scc
; GFX10-NEXT: SI_LOOP [[INT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.2
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.2:
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT]](s32)
; GFX10-NEXT: [[C4:%[0-9]+]]:_(s32) = G_FCONSTANT float 0.000000e+00
; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
; GFX10-NEXT: [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[S_OR_B32_]](s1), [[C5]], [[C4]]
; GFX10-NEXT: G_STORE [[SELECT]](s32), [[MV]](p0) :: (store (s32))
; GFX10-NEXT: SI_RETURN
bb.0:
successors: %bb.1(0x80000000)
liveins: $vgpr0, $vgpr1, $vgpr2
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = COPY $vgpr2
%3:_(p0) = G_MERGE_VALUES %1(s32), %2(s32)
%4:_(s32) = G_CONSTANT i32 0
%5:_(s1) = G_CONSTANT i1 true
bb.1:
successors: %bb.2(0x04000000), %bb.1(0x7c000000)
%6:_(s32) = G_PHI %7(s32), %bb.1, %4(s32), %bb.0
%8:_(s32) = G_PHI %4(s32), %bb.0, %9(s32), %bb.1
%10:_(s1) = G_PHI %5(s1), %bb.0, %11(s1), %bb.1
%12:_(s1) = G_CONSTANT i1 true
%11:_(s1) = G_XOR %10, %12
%13:_(s32) = G_UITOFP %8(s32)
%14:_(s1) = G_FCMP floatpred(ogt), %13(s32), %0
%15:_(s32) = G_CONSTANT i32 1
%9:_(s32) = G_ADD %8, %15
%7:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %14(s1), %6(s32)
SI_LOOP %7(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.2
bb.2:
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %7(s32)
%16:_(s32) = G_FCONSTANT float 0.000000e+00
%17:_(s32) = G_FCONSTANT float 1.000000e+00
%18:_(s32) = G_SELECT %10(s1), %17, %16
G_STORE %18(s32), %3(p0) :: (store (s32))
SI_RETURN
...
---
name: temporal_divergent_i1_non_phi
legalized: true
tracksRegLiveness: true
body: |
; GFX10-LABEL: name: temporal_divergent_i1_non_phi
; GFX10: bb.0:
; GFX10-NEXT: successors: %bb.1(0x80000000)
; GFX10-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[C1:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[DEF:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.1:
; GFX10-NEXT: successors: %bb.2(0x04000000), %bb.1(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI:%[0-9]+]]:sreg_32(s1) = PHI [[DEF]](s1), %bb.0, %19(s1), %bb.1
; GFX10-NEXT: [[PHI1:%[0-9]+]]:_(s32) = G_PHI %7(s32), %bb.1, [[C]](s32), %bb.0
; GFX10-NEXT: [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[C]](s32), %bb.0, %9(s32), %bb.1
; GFX10-NEXT: [[PHI3:%[0-9]+]]:_(s1) = G_PHI [[C1]](s1), %bb.0, %11(s1), %bb.1
; GFX10-NEXT: [[COPY3:%[0-9]+]]:sreg_32(s1) = COPY [[PHI]](s1)
; GFX10-NEXT: [[C2:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[PHI3]], [[C2]]
; GFX10-NEXT: [[COPY4:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI2]](s32)
; GFX10-NEXT: [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[COPY]]
; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI2]], [[C3]]
; GFX10-NEXT: [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[FCMP]](s1), [[PHI1]](s32)
; GFX10-NEXT: [[S_ANDN2_B32_:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY3]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY4]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_]](s1), [[S_AND_B32_]](s1), implicit-def $scc
; GFX10-NEXT: SI_LOOP [[INT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.2
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.2:
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT]](s32)
; GFX10-NEXT: [[C4:%[0-9]+]]:_(s32) = G_FCONSTANT float 0.000000e+00
; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
; GFX10-NEXT: [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[S_OR_B32_]](s1), [[C5]], [[C4]]
; GFX10-NEXT: G_STORE [[SELECT]](s32), [[MV]](p0) :: (store (s32))
; GFX10-NEXT: SI_RETURN
bb.0:
successors: %bb.1(0x80000000)
liveins: $vgpr0, $vgpr1, $vgpr2
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = COPY $vgpr2
%3:_(p0) = G_MERGE_VALUES %1(s32), %2(s32)
%4:_(s32) = G_CONSTANT i32 0
%5:_(s1) = G_CONSTANT i1 true
bb.1:
successors: %bb.2(0x04000000), %bb.1(0x7c000000)
%6:_(s32) = G_PHI %7(s32), %bb.1, %4(s32), %bb.0
%8:_(s32) = G_PHI %4(s32), %bb.0, %9(s32), %bb.1
%10:_(s1) = G_PHI %5(s1), %bb.0, %11(s1), %bb.1
%12:_(s1) = G_CONSTANT i1 true
%11:_(s1) = G_XOR %10, %12
%13:_(s32) = G_UITOFP %8(s32)
%14:_(s1) = G_FCMP floatpred(ogt), %13(s32), %0
%15:_(s32) = G_CONSTANT i32 1
%9:_(s32) = G_ADD %8, %15
%7:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %14(s1), %6(s32)
SI_LOOP %7(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.2
bb.2:
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %7(s32)
%16:_(s32) = G_FCONSTANT float 0.000000e+00
%17:_(s32) = G_FCONSTANT float 1.000000e+00
%18:_(s32) = G_SELECT %11(s1), %17, %16
G_STORE %18(s32), %3(p0) :: (store (s32))
SI_RETURN
...
---
name: loop_with_1break
legalized: true
tracksRegLiveness: true
body: |
; GFX10-LABEL: name: loop_with_1break
; GFX10: bb.0:
; GFX10-NEXT: successors: %bb.1(0x80000000)
; GFX10-NEXT: liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr0
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr1
; GFX10-NEXT: [[MV1:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY3]](s32), [[COPY4]](s32)
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr2
; GFX10-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr3
; GFX10-NEXT: [[MV2:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY5]](s32), [[COPY6]](s32)
; GFX10-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[DEF1:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: [[DEF2:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.1:
; GFX10-NEXT: successors: %bb.3(0x50000000), %bb.5(0x30000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI:%[0-9]+]]:sreg_32(s1) = PHI [[DEF2]](s1), %bb.0, %47(s1), %bb.5
; GFX10-NEXT: [[PHI1:%[0-9]+]]:sreg_32(s1) = PHI [[DEF1]](s1), %bb.0, %37(s1), %bb.5
; GFX10-NEXT: [[PHI2:%[0-9]+]]:_(s32) = G_PHI %13(s32), %bb.5, [[C]](s32), %bb.0
; GFX10-NEXT: [[PHI3:%[0-9]+]]:_(s32) = G_PHI [[C]](s32), %bb.0, %15(s32), %bb.5
; GFX10-NEXT: [[COPY7:%[0-9]+]]:sreg_32(s1) = COPY [[PHI]](s1)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:sreg_32(s1) = COPY [[PHI1]](s1)
; GFX10-NEXT: [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[PHI3]](s32)
; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C1]](s32)
; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[MV1]], [[SHL]](s64)
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32), addrspace 1)
; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[LOAD]](s32), [[C2]]
; GFX10-NEXT: [[C3:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[COPY9:%[0-9]+]]:sreg_32(s1) = COPY [[C3]](s1)
; GFX10-NEXT: [[S_ANDN2_B32_:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY7]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY9]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_]](s1), [[S_AND_B32_]](s1), implicit-def $scc
; GFX10-NEXT: [[COPY10:%[0-9]+]]:sreg_32(s1) = COPY [[S_OR_B32_]](s1)
; GFX10-NEXT: G_BRCOND [[ICMP]](s1), %bb.3
; GFX10-NEXT: G_BR %bb.5
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.2:
; GFX10-NEXT: successors: %bb.4(0x80000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
; GFX10-NEXT: G_STORE [[C4]](s32), [[MV2]](p1) :: (store (s32), addrspace 1)
; GFX10-NEXT: G_BR %bb.4
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.3:
; GFX10-NEXT: successors: %bb.5(0x80000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C5]](s32)
; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[MV]], [[SHL1]](s64)
; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s32), addrspace 1)
; GFX10-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[LOAD1]], [[C6]]
; GFX10-NEXT: G_STORE [[ADD]](s32), [[PTR_ADD1]](p1) :: (store (s32), addrspace 1)
; GFX10-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[PHI3]], [[C6]]
; GFX10-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[PHI3]](s32), [[COPY2]]
; GFX10-NEXT: [[COPY11:%[0-9]+]]:sreg_32(s1) = COPY [[ICMP1]](s1)
; GFX10-NEXT: [[C7:%[0-9]+]]:_(s1) = G_CONSTANT i1 false
; GFX10-NEXT: [[S_ANDN2_B32_1:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY10]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY11]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_1:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_1]](s1), [[S_AND_B32_1]](s1), implicit-def $scc
; GFX10-NEXT: G_BR %bb.5
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.4:
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %34(s32)
; GFX10-NEXT: S_ENDPGM 0
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.5:
; GFX10-NEXT: successors: %bb.6(0x04000000), %bb.1(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI4:%[0-9]+]]:sreg_32(s1) = PHI [[S_OR_B32_]](s1), %bb.1, [[S_OR_B32_1]](s1), %bb.3
; GFX10-NEXT: [[PHI5:%[0-9]+]]:_(s32) = G_PHI [[ADD1]](s32), %bb.3, [[DEF]](s32), %bb.1
; GFX10-NEXT: [[PHI6:%[0-9]+]]:sreg_32_xm0_xexec(s1) = G_PHI [[C7]](s1), %bb.3, [[C3]](s1), %bb.1
; GFX10-NEXT: [[COPY12:%[0-9]+]]:sreg_32(s1) = COPY [[PHI4]](s1)
; GFX10-NEXT: [[COPY13:%[0-9]+]]:sreg_32(s1) = COPY [[PHI6]](s1)
; GFX10-NEXT: [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[COPY12]](s1), [[PHI2]](s32)
; GFX10-NEXT: [[S_ANDN2_B32_2:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY8]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_2:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY13]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_2:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_2]](s1), [[S_AND_B32_2]](s1), implicit-def $scc
; GFX10-NEXT: SI_LOOP [[INT]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.6
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.6:
; GFX10-NEXT: successors: %bb.2(0x40000000), %bb.4(0x40000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT]](s32)
; GFX10-NEXT: [[SI_IF:%[0-9]+]]:sreg_32_xm0_xexec(s32) = SI_IF [[S_OR_B32_2]](s1), %bb.4, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.2
bb.0:
successors: %bb.1(0x80000000)
liveins: $sgpr0, $sgpr1, $sgpr2, $sgpr3, $vgpr0, $vgpr1, $vgpr2
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(p1) = G_MERGE_VALUES %0(s32), %1(s32)
%3:_(s32) = COPY $vgpr2
%4:_(s32) = COPY $sgpr0
%5:_(s32) = COPY $sgpr1
%6:_(p1) = G_MERGE_VALUES %4(s32), %5(s32)
%7:_(s32) = COPY $sgpr2
%8:_(s32) = COPY $sgpr3
%9:_(p1) = G_MERGE_VALUES %7(s32), %8(s32)
%10:_(s32) = G_IMPLICIT_DEF
%11:_(s32) = G_CONSTANT i32 0
bb.1:
successors: %bb.3(0x50000000), %bb.5(0x30000000)
%12:_(s32) = G_PHI %13(s32), %bb.5, %11(s32), %bb.0
%14:_(s32) = G_PHI %11(s32), %bb.0, %15(s32), %bb.5
%16:_(s64) = G_SEXT %14(s32)
%17:_(s32) = G_CONSTANT i32 2
%18:_(s64) = G_SHL %16, %17(s32)
%19:_(p1) = G_PTR_ADD %6, %18(s64)
%20:_(s32) = G_LOAD %19(p1) :: (load (s32), addrspace 1)
%21:_(s32) = G_CONSTANT i32 0
%22:_(s1) = G_ICMP intpred(ne), %20(s32), %21
%23:_(s1) = G_CONSTANT i1 true
G_BRCOND %22(s1), %bb.3
G_BR %bb.5
bb.2:
successors: %bb.4(0x80000000)
%24:_(s32) = G_CONSTANT i32 10
G_STORE %24(s32), %9(p1) :: (store (s32), addrspace 1)
G_BR %bb.4
bb.3:
successors: %bb.5(0x80000000)
%25:_(s32) = G_CONSTANT i32 2
%26:_(s64) = G_SHL %16, %25(s32)
%27:_(p1) = G_PTR_ADD %2, %26(s64)
%28:_(s32) = G_LOAD %27(p1) :: (load (s32), addrspace 1)
%29:_(s32) = G_CONSTANT i32 1
%30:_(s32) = G_ADD %28, %29
G_STORE %30(s32), %27(p1) :: (store (s32), addrspace 1)
%31:_(s32) = G_ADD %14, %29
%32:_(s1) = G_ICMP intpred(ult), %14(s32), %3
%33:_(s1) = G_CONSTANT i1 false
G_BR %bb.5
bb.4:
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %34(s32)
S_ENDPGM 0
bb.5:
successors: %bb.6(0x04000000), %bb.1(0x7c000000)
%15:_(s32) = G_PHI %31(s32), %bb.3, %10(s32), %bb.1
%35:sreg_32_xm0_xexec(s1) = G_PHI %33(s1), %bb.3, %23(s1), %bb.1
%36:_(s1) = G_PHI %32(s1), %bb.3, %23(s1), %bb.1
%13:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %36(s1), %12(s32)
SI_LOOP %13(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.6
bb.6:
successors: %bb.2(0x40000000), %bb.4(0x40000000)
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %13(s32)
%34:sreg_32_xm0_xexec(s32) = SI_IF %35(s1), %bb.4, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.2
...
---
name: nested_loops_temporal_divergence_inner
legalized: true
tracksRegLiveness: true
body: |
; GFX10-LABEL: name: nested_loops_temporal_divergence_inner
; GFX10: bb.0:
; GFX10-NEXT: successors: %bb.1(0x80000000)
; GFX10-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY2]](s32), [[COPY3]](s32)
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX10-NEXT: [[MV1:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
; GFX10-NEXT: [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[COPY]](s32), [[C]]
; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.1:
; GFX10-NEXT: successors: %bb.2(0x80000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI:%[0-9]+]]:_(s32) = G_PHI %12(s32), %bb.3, [[C1]](s32), %bb.0
; GFX10-NEXT: [[PHI1:%[0-9]+]]:_(s32) = G_PHI [[C1]](s32), %bb.0, %14(s32), %bb.3
; GFX10-NEXT: [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[PHI1]](s32)
; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C2]](s32)
; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[MV]], [[SHL]](s64)
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32))
; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[DEF:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: [[COPY6:%[0-9]+]]:sreg_32(s1) = COPY [[FCMP]](s1)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.2:
; GFX10-NEXT: successors: %bb.3(0x04000000), %bb.2(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI2:%[0-9]+]]:sreg_32(s1) = PHI [[COPY6]](s1), %bb.1, %43(s1), %bb.2
; GFX10-NEXT: [[PHI3:%[0-9]+]]:sreg_32(s1) = PHI [[DEF]](s1), %bb.1, %35(s1), %bb.2
; GFX10-NEXT: [[PHI4:%[0-9]+]]:_(s32) = G_PHI [[C3]](s32), %bb.1, %22(s32), %bb.2
; GFX10-NEXT: [[PHI5:%[0-9]+]]:_(s32) = G_PHI %24(s32), %bb.2, [[C3]](s32), %bb.1
; GFX10-NEXT: [[COPY7:%[0-9]+]]:sreg_32(s1) = COPY [[PHI2]](s1)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:sreg_32(s1) = COPY [[PHI3]](s1)
; GFX10-NEXT: [[C4:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[COPY7]], [[C4]]
; GFX10-NEXT: [[COPY9:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI5]](s32)
; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI5]], [[C5]]
; GFX10-NEXT: [[FCMP1:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[LOAD]]
; GFX10-NEXT: [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[FCMP1]](s1), [[PHI4]](s32)
; GFX10-NEXT: [[S_ANDN2_B32_:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY8]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY9]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_]](s1), [[S_AND_B32_]](s1), implicit-def $scc
; GFX10-NEXT: [[COPY10:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: SI_LOOP [[INT]](s32), %bb.2, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.3
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.3:
; GFX10-NEXT: successors: %bb.4(0x04000000), %bb.1(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT]](s32)
; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[MV1]], [[SEXT]](s64)
; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[S_OR_B32_]](s1)
; GFX10-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: G_STORE [[ZEXT]](s32), [[PTR_ADD1]](p0) :: (store (s8))
; GFX10-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[PHI1]], [[C6]]
; GFX10-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[PHI1]](s32), [[COPY1]]
; GFX10-NEXT: [[INT1:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[ICMP]](s1), [[PHI]](s32)
; GFX10-NEXT: SI_LOOP [[INT1]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.4
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.4:
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT1]](s32)
; GFX10-NEXT: SI_RETURN
bb.0:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = COPY $vgpr2
%3:_(s32) = COPY $vgpr3
%4:_(p0) = G_MERGE_VALUES %2(s32), %3(s32)
%5:_(s32) = COPY $vgpr4
%6:_(s32) = COPY $vgpr5
%7:_(p0) = G_MERGE_VALUES %5(s32), %6(s32)
%8:_(s32) = G_FCONSTANT float 1.000000e+00
%9:_(s1) = G_FCMP floatpred(ogt), %0(s32), %8
%10:_(s32) = G_CONSTANT i32 0
bb.1:
%11:_(s32) = G_PHI %12(s32), %bb.3, %10(s32), %bb.0
%13:_(s32) = G_PHI %10(s32), %bb.0, %14(s32), %bb.3
%15:_(s64) = G_SEXT %13(s32)
%16:_(s32) = G_CONSTANT i32 2
%17:_(s64) = G_SHL %15, %16(s32)
%18:_(p0) = G_PTR_ADD %4, %17(s64)
%19:_(s32) = G_LOAD %18(p0) :: (load (s32))
%20:_(s32) = G_CONSTANT i32 0
bb.2:
successors: %bb.3(0x04000000), %bb.2(0x7c000000)
%21:_(s32) = G_PHI %20(s32), %bb.1, %22(s32), %bb.2
%23:_(s32) = G_PHI %24(s32), %bb.2, %20(s32), %bb.1
%25:_(s1) = G_PHI %26(s1), %bb.2, %9(s1), %bb.1
%27:_(s1) = G_CONSTANT i1 true
%26:_(s1) = G_XOR %25, %27
%28:_(s32) = G_UITOFP %23(s32)
%29:_(s32) = G_CONSTANT i32 1
%24:_(s32) = G_ADD %23, %29
%30:_(s1) = G_FCMP floatpred(ogt), %28(s32), %19
%22:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %30(s1), %21(s32)
SI_LOOP %22(s32), %bb.2, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.3
bb.3:
successors: %bb.4(0x04000000), %bb.1(0x7c000000)
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %22(s32)
%31:_(p0) = G_PTR_ADD %7, %15(s64)
%32:_(s32) = G_ZEXT %26(s1)
%33:_(s32) = G_CONSTANT i32 1
G_STORE %32(s32), %31(p0) :: (store (s8))
%14:_(s32) = G_ADD %13, %33
%34:_(s1) = G_ICMP intpred(ult), %13(s32), %1
%12:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %34(s1), %11(s32)
SI_LOOP %12(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.4
bb.4:
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %12(s32)
SI_RETURN
...
---
name: nested_loops_temporal_divergence_outer
legalized: true
tracksRegLiveness: true
body: |
; GFX10-LABEL: name: nested_loops_temporal_divergence_outer
; GFX10: bb.0:
; GFX10-NEXT: successors: %bb.1(0x80000000)
; GFX10-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY2]](s32), [[COPY3]](s32)
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX10-NEXT: [[MV1:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
; GFX10-NEXT: [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[COPY]](s32), [[C]]
; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.1:
; GFX10-NEXT: successors: %bb.2(0x80000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI:%[0-9]+]]:_(s32) = G_PHI %12(s32), %bb.3, [[C1]](s32), %bb.0
; GFX10-NEXT: [[PHI1:%[0-9]+]]:_(s32) = G_PHI [[C1]](s32), %bb.0, %14(s32), %bb.3
; GFX10-NEXT: [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[PHI1]](s32)
; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C2]](s32)
; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[MV]], [[SHL]](s64)
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32))
; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[DEF:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: [[COPY6:%[0-9]+]]:sreg_32(s1) = COPY [[FCMP]](s1)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.2:
; GFX10-NEXT: successors: %bb.3(0x04000000), %bb.2(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI2:%[0-9]+]]:sreg_32(s1) = PHI [[COPY6]](s1), %bb.1, %43(s1), %bb.2
; GFX10-NEXT: [[PHI3:%[0-9]+]]:sreg_32(s1) = PHI [[DEF]](s1), %bb.1, %35(s1), %bb.2
; GFX10-NEXT: [[PHI4:%[0-9]+]]:_(s32) = G_PHI [[C3]](s32), %bb.1, %22(s32), %bb.2
; GFX10-NEXT: [[PHI5:%[0-9]+]]:_(s32) = G_PHI %24(s32), %bb.2, [[C3]](s32), %bb.1
; GFX10-NEXT: [[COPY7:%[0-9]+]]:sreg_32(s1) = COPY [[PHI2]](s1)
; GFX10-NEXT: [[COPY8:%[0-9]+]]:sreg_32(s1) = COPY [[PHI3]](s1)
; GFX10-NEXT: [[C4:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[COPY7]], [[C4]]
; GFX10-NEXT: [[COPY9:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI5]](s32)
; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI5]], [[C5]]
; GFX10-NEXT: [[FCMP1:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[LOAD]]
; GFX10-NEXT: [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[FCMP1]](s1), [[PHI4]](s32)
; GFX10-NEXT: [[S_ANDN2_B32_:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY8]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY9]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_]](s1), [[S_AND_B32_]](s1), implicit-def $scc
; GFX10-NEXT: [[COPY10:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: SI_LOOP [[INT]](s32), %bb.2, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.3
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.3:
; GFX10-NEXT: successors: %bb.4(0x04000000), %bb.1(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT]](s32)
; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[MV1]], [[SEXT]](s64)
; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[S_OR_B32_]](s1)
; GFX10-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: G_STORE [[ZEXT]](s32), [[PTR_ADD1]](p0) :: (store (s8))
; GFX10-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[PHI1]], [[C6]]
; GFX10-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[PHI1]](s32), [[COPY1]]
; GFX10-NEXT: [[INT1:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[ICMP]](s1), [[PHI]](s32)
; GFX10-NEXT: SI_LOOP [[INT1]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.4
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.4:
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT1]](s32)
; GFX10-NEXT: SI_RETURN
bb.0:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = COPY $vgpr2
%3:_(s32) = COPY $vgpr3
%4:_(p0) = G_MERGE_VALUES %2(s32), %3(s32)
%5:_(s32) = COPY $vgpr4
%6:_(s32) = COPY $vgpr5
%7:_(p0) = G_MERGE_VALUES %5(s32), %6(s32)
%8:_(s32) = G_FCONSTANT float 1.000000e+00
%9:_(s1) = G_FCMP floatpred(ogt), %0(s32), %8
%10:_(s32) = G_CONSTANT i32 0
bb.1:
%11:_(s32) = G_PHI %12(s32), %bb.3, %10(s32), %bb.0
%13:_(s32) = G_PHI %10(s32), %bb.0, %14(s32), %bb.3
%15:_(s64) = G_SEXT %13(s32)
%16:_(s32) = G_CONSTANT i32 2
%17:_(s64) = G_SHL %15, %16(s32)
%18:_(p0) = G_PTR_ADD %4, %17(s64)
%19:_(s32) = G_LOAD %18(p0) :: (load (s32))
%20:_(s32) = G_CONSTANT i32 0
bb.2:
successors: %bb.3(0x04000000), %bb.2(0x7c000000)
%21:_(s32) = G_PHI %20(s32), %bb.1, %22(s32), %bb.2
%23:_(s32) = G_PHI %24(s32), %bb.2, %20(s32), %bb.1
%25:_(s1) = G_PHI %26(s1), %bb.2, %9(s1), %bb.1
%27:_(s1) = G_CONSTANT i1 true
%26:_(s1) = G_XOR %25, %27
%28:_(s32) = G_UITOFP %23(s32)
%29:_(s32) = G_CONSTANT i32 1
%24:_(s32) = G_ADD %23, %29
%30:_(s1) = G_FCMP floatpred(ogt), %28(s32), %19
%22:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %30(s1), %21(s32)
SI_LOOP %22(s32), %bb.2, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.3
bb.3:
successors: %bb.4(0x04000000), %bb.1(0x7c000000)
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %22(s32)
%31:_(p0) = G_PTR_ADD %7, %15(s64)
%32:_(s32) = G_ZEXT %26(s1)
%33:_(s32) = G_CONSTANT i32 1
G_STORE %32(s32), %31(p0) :: (store (s8))
%14:_(s32) = G_ADD %13, %33
%34:_(s1) = G_ICMP intpred(ult), %13(s32), %1
%12:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %34(s1), %11(s32)
SI_LOOP %12(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.4
bb.4:
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %12(s32)
SI_RETURN
...
---
name: nested_loops_temporal_divergence_both
legalized: true
tracksRegLiveness: true
body: |
; GFX10-LABEL: name: nested_loops_temporal_divergence_both
; GFX10: bb.0:
; GFX10-NEXT: successors: %bb.1(0x80000000)
; GFX10-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr3
; GFX10-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY2]](s32), [[COPY3]](s32)
; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr4
; GFX10-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr5
; GFX10-NEXT: [[MV1:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
; GFX10-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY $vgpr6
; GFX10-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY $vgpr7
; GFX10-NEXT: [[MV2:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY6]](s32), [[COPY7]](s32)
; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
; GFX10-NEXT: [[FCMP:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[COPY]](s32), [[C]]
; GFX10-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[DEF:%[0-9]+]]:sreg_32(s1) = IMPLICIT_DEF
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.1:
; GFX10-NEXT: successors: %bb.2(0x80000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI:%[0-9]+]]:sreg_32(s1) = PHI [[DEF]](s1), %bb.0, %39(s1), %bb.3
; GFX10-NEXT: [[PHI1:%[0-9]+]]:_(s32) = G_PHI %15(s32), %bb.3, [[C1]](s32), %bb.0
; GFX10-NEXT: [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[C1]](s32), %bb.0, %17(s32), %bb.3
; GFX10-NEXT: [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[PHI2]](s32)
; GFX10-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
; GFX10-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[SEXT]], [[C2]](s32)
; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[MV]], [[SHL]](s64)
; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p0) :: (load (s32))
; GFX10-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; GFX10-NEXT: [[COPY8:%[0-9]+]]:sreg_32(s1) = COPY [[FCMP]](s1)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.2:
; GFX10-NEXT: successors: %bb.3(0x04000000), %bb.2(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: [[PHI3:%[0-9]+]]:sreg_32(s1) = PHI [[COPY8]](s1), %bb.1, %48(s1), %bb.2
; GFX10-NEXT: [[PHI4:%[0-9]+]]:sreg_32(s1) = PHI [[PHI]](s1), %bb.1, %39(s1), %bb.2
; GFX10-NEXT: [[PHI5:%[0-9]+]]:_(s32) = G_PHI [[C3]](s32), %bb.1, %25(s32), %bb.2
; GFX10-NEXT: [[PHI6:%[0-9]+]]:_(s32) = G_PHI %27(s32), %bb.2, [[C3]](s32), %bb.1
; GFX10-NEXT: [[COPY9:%[0-9]+]]:sreg_32(s1) = COPY [[PHI3]](s1)
; GFX10-NEXT: [[COPY10:%[0-9]+]]:sreg_32(s1) = COPY [[PHI4]](s1)
; GFX10-NEXT: [[C4:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
; GFX10-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[COPY9]], [[C4]]
; GFX10-NEXT: [[COPY11:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: [[UITOFP:%[0-9]+]]:_(s32) = G_UITOFP [[PHI6]](s32)
; GFX10-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[PHI6]], [[C5]]
; GFX10-NEXT: [[FCMP1:%[0-9]+]]:_(s1) = G_FCMP floatpred(ogt), [[UITOFP]](s32), [[LOAD]]
; GFX10-NEXT: [[INT:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[FCMP1]](s1), [[PHI5]](s32)
; GFX10-NEXT: [[S_ANDN2_B32_:%[0-9]+]]:sreg_32(s1) = S_ANDN2_B32 [[COPY10]](s1), $exec_lo, implicit-def $scc
; GFX10-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32(s1) = S_AND_B32 $exec_lo, [[COPY11]](s1), implicit-def $scc
; GFX10-NEXT: [[S_OR_B32_:%[0-9]+]]:sreg_32(s1) = S_OR_B32 [[S_ANDN2_B32_]](s1), [[S_AND_B32_]](s1), implicit-def $scc
; GFX10-NEXT: [[COPY12:%[0-9]+]]:sreg_32(s1) = COPY [[XOR]](s1)
; GFX10-NEXT: SI_LOOP [[INT]](s32), %bb.2, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.3
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.3:
; GFX10-NEXT: successors: %bb.4(0x04000000), %bb.1(0x7c000000)
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT]](s32)
; GFX10-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[MV1]], [[SEXT]](s64)
; GFX10-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[S_OR_B32_]](s1)
; GFX10-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
; GFX10-NEXT: G_STORE [[ZEXT]](s32), [[PTR_ADD1]](p0) :: (store (s8))
; GFX10-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[PHI2]], [[C6]]
; GFX10-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ult), [[PHI2]](s32), [[COPY1]]
; GFX10-NEXT: [[INT1:%[0-9]+]]:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), [[ICMP]](s1), [[PHI1]](s32)
; GFX10-NEXT: SI_LOOP [[INT1]](s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
; GFX10-NEXT: G_BR %bb.4
; GFX10-NEXT: {{ $}}
; GFX10-NEXT: bb.4:
; GFX10-NEXT: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), [[INT1]](s32)
; GFX10-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[S_OR_B32_]](s1)
; GFX10-NEXT: G_STORE [[ZEXT1]](s32), [[MV2]](p0) :: (store (s8))
; GFX10-NEXT: SI_RETURN
bb.0:
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = COPY $vgpr2
%3:_(s32) = COPY $vgpr3
%4:_(p0) = G_MERGE_VALUES %2(s32), %3(s32)
%5:_(s32) = COPY $vgpr4
%6:_(s32) = COPY $vgpr5
%7:_(p0) = G_MERGE_VALUES %5(s32), %6(s32)
%8:_(s32) = COPY $vgpr6
%9:_(s32) = COPY $vgpr7
%10:_(p0) = G_MERGE_VALUES %8(s32), %9(s32)
%11:_(s32) = G_FCONSTANT float 1.000000e+00
%12:_(s1) = G_FCMP floatpred(ogt), %0(s32), %11
%13:_(s32) = G_CONSTANT i32 0
bb.1:
%14:_(s32) = G_PHI %15(s32), %bb.3, %13(s32), %bb.0
%16:_(s32) = G_PHI %13(s32), %bb.0, %17(s32), %bb.3
%18:_(s64) = G_SEXT %16(s32)
%19:_(s32) = G_CONSTANT i32 2
%20:_(s64) = G_SHL %18, %19(s32)
%21:_(p0) = G_PTR_ADD %4, %20(s64)
%22:_(s32) = G_LOAD %21(p0) :: (load (s32))
%23:_(s32) = G_CONSTANT i32 0
bb.2:
successors: %bb.3(0x04000000), %bb.2(0x7c000000)
%24:_(s32) = G_PHI %23(s32), %bb.1, %25(s32), %bb.2
%26:_(s32) = G_PHI %27(s32), %bb.2, %23(s32), %bb.1
%28:_(s1) = G_PHI %29(s1), %bb.2, %12(s1), %bb.1
%30:_(s1) = G_CONSTANT i1 true
%29:_(s1) = G_XOR %28, %30
%31:_(s32) = G_UITOFP %26(s32)
%32:_(s32) = G_CONSTANT i32 1
%27:_(s32) = G_ADD %26, %32
%33:_(s1) = G_FCMP floatpred(ogt), %31(s32), %22
%25:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %33(s1), %24(s32)
SI_LOOP %25(s32), %bb.2, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.3
bb.3:
successors: %bb.4(0x04000000), %bb.1(0x7c000000)
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %25(s32)
%34:_(p0) = G_PTR_ADD %7, %18(s64)
%35:_(s32) = G_ZEXT %29(s1)
%36:_(s32) = G_CONSTANT i32 1
G_STORE %35(s32), %34(p0) :: (store (s8))
%17:_(s32) = G_ADD %16, %36
%37:_(s1) = G_ICMP intpred(ult), %16(s32), %1
%15:sreg_32_xm0_xexec(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.if.break), %37(s1), %14(s32)
SI_LOOP %15(s32), %bb.1, implicit-def $exec, implicit-def $scc, implicit $exec
G_BR %bb.4
bb.4:
G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.end.cf), %15(s32)
%38:_(s32) = G_ZEXT %29(s1)
G_STORE %38(s32), %10(p0) :: (store (s8))
SI_RETURN
...
Reference in new issue View Git Blame Copy Permalink